What's new since v1.1.2

get_story: optional include_notes parameter

Set include_notes=false to omit the ## Notes section (and everything after it) from the returned content. Story files only grow over time — every add_story_note and complete_story call appends a timestamped note, and nothing ever trims it — so a heavily-annotated, long-lived story can carry a lot of accumulated history that a caller doesn't always need. Defaults to true; existing callers see no change.

bulk_update_acceptance_criteria no longer echoes file content

The response no longer includes the full post-update file content — it now returns {story_id, path, criteria_updated, errors}. The criteria_updated/errors pair was already sufficient signal; callers who want to see the resulting file should call get_story explicitly. Both of these changes were flagged from a real dogfood session where repeated calls made up a disproportionate share of the session's token spend.

Bug fix: bare AC IDs with non-hex suffixes now resolve correctly

bulk_update_acceptance_criteria and check_acceptance_criterion previously only recognised AC IDs with an 8-character hex suffix (e.g. AC-STORY-042-a3f9b2c1). Stories whose acceptance criteria were hand-authored with plain sequential IDs instead (e.g. AC-STORY-159-1, AC-STORY-159-2, ...) — most commonly when an external tool or agent writes AC lines directly to markdown rather than going through set_acceptance_criteria — would fail lookup for every criterion with "criterion/criteria not found", even though the same call worked fine for hex-suffixed stories. The ID regex now accepts any-length hex suffix, so both formats resolve.

What's fixed

set_acceptance_criteria no longer corrupts acceptance criteria when passed pre-ticked or fully-formatted input.

Previously, set_acceptance_criteria unconditionally prepended - [ ] to every criterion string, even when the string already began with a checkbox marker ([x] ..., - [x] ...) or a full stored line including an existing AC-ID. This produced double-nested, broken checkboxes (e.g. - [ ] [x] AC-STORY-114-abc: text), and the ID-assignment logic would then treat the original ID as ordinary text and generate a brand new one — silently breaking any subsequent tool call that referenced the old ID. Recovery previously required a direct file edit.

set_acceptance_criteria now strips any leading checkbox marker before processing, preserves the ticked state of the input, and keeps existing AC-IDs intact rather than regenerating them. This makes it safe to re-pass criteria lines read directly from a story file (e.g. during bulk updates or story review), which is the natural workflow for an agent working with get_story output.

Also in this release

• Filed two new bug stories (STORY-032, STORY-033) in the backlog covering list_stories Unicode mojibake in story titles and a bulk_update_acceptance_criteria key-lookup failure on combined ID: text inputs. Both are tracked for future fixes.

Upgrading

brew upgrade backlog-mcp

or download the binary for your platform from the assets below.

Changelog

• a6f62c4 build: add x86 (386) build targets for Windows and Linux

What's new since v1.0.4

New tool: reorder_backlog

Reorder the active backlog by supplying a priority-ordered list of story IDs. Stories omitted from the list are appended at the end — nothing is silently dropped. Uses the same atomic write + file lock path as all other mutating tools.

Bug fixes

• Acceptance criteria text matching — check_acceptance_criterion and bulk_update_acceptance_criteria now tolerate Unicode dash variants (em-dash, en-dash, figure dash, etc.) in criterion text, so an agent that stored a criterion with — can look it up with – or - without a criteria_errors failure.
• RemoveFromBacklog now reports whether an entry was actually removed — previously it silently returned nil even if the story wasn't in the backlog.
• bulk_update_stories surfaces backlog update failures — previously, if a story's backlog entry was missing during a status update, the error was silently discarded. It now appears in the per-story errors array.

Discoverability

• list_stories description now lists all other tools in the server, so agents in deferred-tool environments (e.g. VS Code Copilot) discover the full tool surface from a single lookup.
• bulk_update_stories criteria format is now documented with a concrete example in the tool description.

Docs

• README updated with Homebrew install instructions and an expanded tool list.

What's Changed

• chore: fix backlog-agent install action to use go install instead of go build by @Copilot in #5
• Copilot/story 022 add bulk update tools by @corbym in #7
• feat: lazy acceptance criteria ID generation by @Copilot in #8
• feat(STORY-023): Add Homebrew tap formula and release automation by @Copilot in #9
• Allow Release workflow to dispatch Homebrew update workflow by @Copilot in #10
• chore: replace manual Homebrew tap workflow with goreleaser native brews support by @Copilot in #11

Full Changelog: v1.0.3...v1.0.4

Changelog

• 459eb58 feat: add FindEpicFilePath function to retrieve relative path of epic markdown files
• f5d6c7f feat: add set_epic_status tool to manage epic lifecycle and update requirements index
• de82e09 feat: update README with new epic status management details and guards

Added epic status change command to manage epic lifecycle and update requirements index

Changelog

• 02a2768 feat: add set_epic_status tool to manage epic lifecycle and update requirements index

Full Changelog: v1.0.1...v1.0.2

Changelog

• 7328b87 Merge pull request #3 from corbym/copilot/promote-release-v016-to-v100
• 7400a32 Merge pull request #4 from corbym/copilot/update-readme-http-port-info
• 601f6ba feat: add promote-release workflow for tagging aliases without rebuild
• a4e8d57 feat: enhance story status management and update related configurations
• 3ace4b6 fix CLAUDE.md: remove stale HTTP refs, add plan subcommand, fix repo layout
• 20afc8a fix README: remove unimplemented HTTP env vars, add plan subcommand docs
• 3b3e631 fix: update acceptance criteria descriptions and improve error messaging in tools

Changelog (v0.1.4 → v0.1.6)

Comparison: v0.1.4...v0.1.6

Highlights

• Expanded the MCP server from 5 tools → 9 tools, covering full backlog lifecycle: create epics, create stories, set acceptance criteria, and complete stories (with validation), plus new maintenance tooling.
• Added file locking for all mutating operations to prevent concurrent-write corruption.
• Introduced story types (feature/bug/chore/spike) end-to-end (create, index storage, list/get, filtering).
• Improved backlog and epic navigation by adding relative links and maintaining an epic “Stories” section.
• Added acceptance-criteria parsing and enforcement, including a dedicated tool to check off criteria line-by-line.
• Added a groom_epic tool to reconcile epic story lists with the filesystem and index.

Added / Changed MCP Tools

New tools (added since v0.1.4)

• create_epic
  • Creates an epic directory + epic.md, assigns next EPIC-NNN, and registers in requirements-index.md.
• create_story
  • Creates a story under an epic, assigns next STORY-NNN, writes story markdown, registers in index + backlog.
  • Now supports story_type.
• set_acceptance_criteria
  • Replaces the story’s “Acceptance criteria” section from an array of strings.
  • Schema bug fixed: the criteria array is now properly declared/required in the tool schema.
• complete_story
  • One-call completion: validates AC, marks done in index, removes from backlog, appends completion note.
  • Supports incomplete_items when criteria are unchecked.
• check_acceptance_criterion
  • Checks a single AC item (- [ ] → - [x]) by 0-based index or case-insensitive text match.
• groom_epic
  • Reconciles an epic’s ## Stories section against story files on disk + index metadata.

Updated existing tools / behavior changes

• set_story_status
  • Now only allows: draft, in-progress, blocked.
  • Setting status to done is blocked and returns an error instructing callers to use complete_story.
  • If a story isn’t found in backlog.md, the tool now returns a backlog_warning instead of silently ignoring the failure.
• list_stories
  • Now supports filtering by story_type and returns story_type in results.
• get_story
  • Now includes story_type in the returned metadata.

Concurrency & Safety

• File locking added for mutating tools via parser.AcquireLock(...) (5-second timeout), with platform-specific implementations:
  • parser/lock_unix.go
  • parser/lock_windows.go
• Reinforced “atomic writes everywhere” pattern (still uses atomic write strategy).

Backlog + Epic Markdown UX Improvements

• Backlog entries are now link-based, supporting navigation:
  • New format: N. [STORY-NNN](story-path) ([EPIC-NNN](epic-path)) — title *(status)*
  • Parser updated to support both old and new formats, including safer renumbering and status updates.
• Epic files now maintain a ## Stories section
  • create_story updates the epic file with a checklist entry for the story.
  • complete_story flips the epic checklist entry to done ([x]) (best-effort / non-fatal).
• groom_epic tool can rebuild/sync that ## Stories section:
  • Adds missing stories, removes stale entries, refreshes titles and checked-state from index, keeps entries ordered.

Story “Type” Support (feature/bug/chore/spike)

• create_story accepts optional story_type (defaults to feature, validates allowed values).
• New stories include a **Type:** ... line in the story file.
• requirements-index.md epic tables were extended to include an optional Type column (with backward compat parsing).

Acceptance Criteria: Parsing, Enforcement, and Completion Gating

• Added parsing utilities:
  • ParseAcceptanceCriteria reads ## Acceptance criteria checklist items into structured {Text, Checked}.
• complete_story now enforces:
  • Blocks completion if AC is placeholder-only (“Define acceptance criteria”).
  • If any unchecked criteria exist, requires incomplete_items with one explanation per unchecked item, included in the completion note.
• New check_acceptance_criterion tool implements line-level checklist progress.

Tests & Quality

• Added a substantial set of tests covering:
  • Grooming epic story sections (parser/groom_test.go)
  • Acceptance-criteria gating behavior in complete_story
  • check_acceptance_criterion behaviors (index/text/case-insensitive/already checked/out-of-range/not found/etc.)
  • set_story_status redirect-from-done behavior and missing-backlog warning behavior
• Updated fixtures to include additional epic/story files and acceptance criteria states.

Documentation / Metadata

• README.md: added a Glama MCP server badge.
• CLAUDE.md: updated to reflect 9 tools (was 5), added detailed tool docs, and updated design decisions to reflect file locking.
• Some IDE/project files were added/changed (e.g., .idea/editor.xml, backlog.iml) as part of development environment configuration.

Notable Internal Implementation Changes (developer-facing)

• parser/backlog.go: parsing and renumbering rewritten to support the new link-style backlog format.
• parser/create.go: CreateStory signature changed to accept storyType; new helpers to update backlog + epic files.
• parser/index.go: index parsing now supports an optional “Type” column and defaults legacy rows to feature.
• parser/story.go: acceptance-criteria parsing plus new “check one criterion” mutation function.

Full Changelog: v0.1.4...v0.1.6

Changelog

• 432c5ac Merge pull request #2 from corbym/copilot/check-release-build-failure
• fb80bc5 fix: split lock.go into lock_unix.go and lock_windows.go for cross-platform build